Some thoughts about the 1940, 1944, and 1948 elections

In several posts (such as this one, this one, and this one), it is established that 1948 was different to 1940 and 1944 other than simply being closer than those two. 

The most obvious sign of this is that Truman carried four states that Dewey had carried in 1944: Colorado, Wyoming, Iowa, and Wisconsin. Had Dewey won the Electoral College in 1948 by carrying Illinois, Ohio, and California (which were all closer in 1948 than any of Truman's four turnover states), there would have been states voting for losers of two different major parties in consecutive elections for the first time since 1896. (As it happened, 1896 remains the last time that any state has voted for losers of two different major parties in consecutive elections.)

The largest explanation of this seems to be isolationism. As Kevin Phillips wrote in The Emerging Republican Majority (p. 462-463),

Then in 1940, the lure of isolationism beckoned many voters...the Republican vote grew large enough for presidential nominee Wendell Willkie to carry Colorado and just miss capturing Wyoming. In 1944, Wyoming also turned against Roosevelt. The isolationist GOP trend was most emphatic in the plains sections of Wyoming, Montana, and Colorado, where 1936-40 Republican gains were commonly in the 15 per cent to 20 per cent range.

...

But what isolationism had joined together, isolationism subsequently pulled asunder. Many Rocky Mountain isolationist voters of 1940 and 1944 stayed home in 1948, destroying the wartime-inflated GOP majorities in Colorado and Wyoming. Shorn of this Republican strength, the Mountain states voted Democratic in 1948. Truman carried all eight. Dewey's loss of Colorado and Wyoming (he was the candidate both states had supported in 1944) was similar in cause to his loss of Iowa, Wisconsin, and Ohio.

Some pages before (on p. 458), he had differentiated the Mountain West into two sub-regions:

Thus, the Mountain states are much more geopolitically complex than laymen are wont to think. A rough but useful division can be made between the Mountain-Plains states (Colorado, Montana, and Wyoming) with their wheat and cattle; their Germans, Scandinavians, and Irish; their isolationists; their agrarian, mine, mill and smelter radicals, and the interior plateau states (Arizona, New Mexico, Nevada, Utah, and Idaho) with their booming resorts and retirement towns; their cotton and truck farming; their deserts; their Mormons, Mexicans, Navajos, Apaches, and Southern Democratic traditionalists.

(Leaving aside that Phillips is talking only of the interior Mountain West, the division he makes is roughly the same as that accomplished by distinguishing the 'Pacific Slope states'--the states lying west of the Continental Divide--from the rest of the Mountain West.) 

The examples of Wyoming and Colorado would militate that 1948 was a more 'normal' election, featuring more normal coalitions for each of the two major nominees. Colorado had voted for William Jennings Bryan thrice (one of only two states outside the South to do so, along with Nevada), and was Wilson's best state outside the South in 1916. Wyoming had voted for Bryan in 1896 (albeit not at similar levels as the other Mountain states), and had also voted for Wilson twice. In 1940, Roosevelt became the first Democrat to win without Colorado since 1892, and in 1944, he became the first to win without Wyoming (as well as the second to win without Colorado) since 1892. In 1948, Truman became the first Democrat to win without Maryland since 1856--but, had Dewey himself won without carrying Maryland, he would have become the first Republican to win without Maryland since 1888. Aside from that, every state Dewey carried in 1948 had either voted for Hughes in 1916, or voted for Hoover in 1932, or voted for both Willkie in 1940 and Dewey in 1944.

Speaking of Hoover in 1932, the examples of Pennsylvania, Delaware, Connecticut, and New Hampshire would also suggest that 1948 was a more 'normal' election than 1940 and 1944. These were four of the six states Hoover carried in his landslide defeat in 1932. As both Willkie in 1940 and Dewey in 1944 outperformed Hoover's '32 performance, one would expect them to be able to reclaim these states (all of which save New Hampshire had also voted for Hughes in 1916). But they did not. Only in 1948 was Dewey able to finally bring them all back into the Republican fold.

Further indicating the greater 'normality' of 1948 vis-à-vis 1940 and 1944 is the far greater number of counties Roosevelt was the first Democrat in the 20th century to win without in 1940 and 1944, than Truman was in 1948 (excluding Thurmond counties). Furthermore, 52 of the counties FDR was the first Democrat in the 20th century to win without in 1940 went on to vote for Dewey in 1944 and Truman in 1948; and 47 of the counties FDR was the first Democrat in the 20th century to win without in 1944 flipped back to Truman in 1948. 

However, it is not as simple as that. Two of the Dewey '44-Truman '48 states--Iowa and Wisconsin--were historically Republican states as of 1948, and would remain so for some decades thereafter. Truman won both with an outright majority (50.70% in Wisconsin and 50.31% in Iowa). Since Wisconsin's first participating election (in 1848), 1948 was only the fifth time the state had given the Democrat a majority, after 1852, 1932, 1936, and 1940. Since Iowa's first participating election (also in 1848), 1948 was only the fourth time that state had given the Democrat a majority, after 1852, 1932, and 1936. In both cases, it was the first time ever that either state had ever given a majority to the Democrat even as the Democrat was not winning a national majority. (Both states had also voted for Hughes in 1916.)

After 1948, Wisconsin voted Republican in 1960 and 1968, but narrowly voted for Carter in 1976. As Carter won only a plurality in the state, 1948 would remain the solitary Democratic majority in Wisconsin (apart from 1964) between 1940 and 1988. Iowa voted for Nixon in 1960, Nixon in 1968, and Ford in 1976; 1948 would (again, apart from 1964) stand, not only as the sole Democratic majority in Iowa, but also as the only time the Democrats even carried Iowa, between 1936 and 1988.

Had Dewey won the Electoral College in 1948 by carrying California, Illinois, and Ohio, he would have become the first Republican ever to win without Iowa (a distinction that instead went to George H. W. Bush). He would have become the first Republican to win despite losing Wisconsin to the Democrat (Coolidge had won in 1924 while losing Wisconsin to Progressive nominee Robert La Follette)--this distinction, again, instead went to George H. W. Bush. 

This is a map of the states coloured in by whether Dewey's vote share rose (red) or declined (blue) in 1948 relative to 1944. Helping explain how Dewey offset his losses in certain parts of the country, he increased his vote share in three of the four largest states: Pennsylvania, Illinois, and California. (California and Ohio both had 25 electoral votes in 1948, but California cast more votes.) States where he improved his vote share by over 2% were:

Texas (+7.65%)
Utah (+5.60%)
Delaware (+4.77%)
New Hampshire (+4.54%)
Vermont (+4.48%)
Maine (+4.30%)
California (+4.19%)
Florida (+3.95%)
Virginia (+3.65%)
Arizona (+2.92%)
Oregon (+2.84%)
Connecticut (+2.61%)
Pennsylvania (+2.57%)

So, while Dewey lost ground in Colorado, Wyoming, and Montana (and New Mexico), he not only gained almost across the board in the Pacific Slope states, but gained significantly. His improvement in Utah was his second-best in the country, and the already-large state of California, was one of only seven in which he improved by over 4%. (In Nevada, he improved by 1.88%.)

He also improved substantially in three former Confederate states--Virginia, Texas (home to his best improvement in the country), and Florida. According to Howard K. Smith in 1976, these were also the three states that had been 'denominated by Jimmy Carter as...his three vital battlegrounds in the South', implying that the Carter campaign saw them as shakier than other Southern states. And, as it turned out, Texas and Florida were, apart from Mississippi (which had last voted Democratic in 1956), Carter's closest former Confederate states; and Virginia was the only former Confederate state that actually voted for Ford. 

These were arguably the first three former Confederate states to join what would become the 'Sun Belt'. (In Maryland--where Dewey had improved on Willkie in 1944 by 7.32%--Dewey improved by 1.75% in 1948. His improvements in Alabama and Georgia were both under 1%.)

In fact, Phillips directly ties improvements in Dewey's vote share in the Mountain West to whether a given state was growing or not:

The same isolationism which figured so obviously in the Great Plains-influenced Mountain states was relatively unimportant in the Southwest, where Dewey did better in 1948 than he had in 1944. Whereas the number of presidential votes cast declined between 1940 and 1948 in Wyoming, Colorado, Montana, and Idaho, it increased in Arizona, New Mexico, Nevada, and Utah. The Southwest boomed after World War II, and many of the immigrants were Northern Republicans. [p. 463]

And it seems this was likely true in the South (and relevant to California) as well, especially when one looks at some of the Davis '24-Dewey '48 counties in the South (there were scarcely any outside the South): Broward, Orange, and Sarasota, FL; Montgomery, MD; Fairfax, VA. These counties cast 1,887.9%, 524.4%, 1,442.2%, 197.9%, and 272.1% more votes in 1948 than in 1924, respectively (the country overall cast 67.7% more votes).

Particularly in the Tidewater states of Virginia and Maryland, there were also rural counties that were unlikely to be growing particularly fast at the time, that were also Davis '24-Dewey '48 counties (such as Caroline and Talbot, MD and King George, Botetourt, Mathews, and Northumberland, VA). 

In Texas, there were only two Davis '24-Dewey '48 counties, and they were small (Washington and Austin Counties, the latter of which is not home to the city of Austin). However, in Harris, Dallas, and Tarrant Counties--the largest, second-largest, and fourth-largest counties in the state (and all soon to become Republican strongholds)--he did over ten points better, in 1948, than any Republican had managed to do against FDR in any of his elections (for some reason, the Republican high-water mark in each of these counties, in the previous four elections, had been reached in 1940). He improved on his own 1944 vote share by over 23% in Harris County, by over 15% in Dallas County, and by over 20% in Tarrant County. He still fell well short of carrying any of them (his best vote share in any of them in 1948 was 37.80%, in Dallas County), but it seems likely his improvements in these counties were largely what powered his overall substantial improvement in Texas.

Ironically, the nature of Dewey's improvements largely meant that they were wasted in the Electoral College. His improvement in California, while impressive, was not enough to flip the state (which had staunchly backed FDR in all four of his elections). As Truman took back Colorado and Wyoming, Dewey's improvements were not enough to break into the interior Pacific Slope states, which had all backed every Democratic winner in the 20th century (and several of which had been amongst FDR's best states outside the South in his elections). The Solid South states of Florida and Virginia held for Truman fairly comfortably (he won Virginia by over 6% and Florida by over 15%)--and as for Texas, it was his best state in the country. Texas and Utah--the two states in which Dewey improved his vote share by over 5%--voted for Truman by 41.7% and 8.96%, respectively.

---

What does this mean for the respective desirability of using 1940, 1944, and 1948 as a comparison-point election? 

Well, although in some ways 1948 seems more 'normal' than 1940 and 1944, in other ways it wasn't. The removal of isolationism as a motivating factor might explain Colorado and Wyoming, but, given their Republican pasts (and Republican futures), it, on its own, isn't enough to explain Iowa and Wisconsin.

1940 might be the best comparison-point election (other than that it makes the elapsed time between itself and whatever election we're comparing it with greater). It isn't itself free of oddities--for example, a list of Willkie-Kerry counties would include Dubuque County, Iowa (and although many Midwestern Dewey '44-Kerry counties were counties that had been traditionally Republican before turning Democratic sometime during the Cold War period, before once again turning Republican sometime in the last few elections, Dubuque was not one of them--in 1940, FDR became the first Democrat to win despite losing the county since it became a county ahead of the 1848 election; the second would be Joe Biden in 2020). 

But Willkie did badly enough--losing the national popular vote by 9.96%, and getting just 44.78% of the vote--that, regardless of any oddities at work in that election, any subsequent nationally competitive Republican should have been able to hold whatever he had won--barring some sort of realignment that was allowing this hypothetical future candidate to make up for it. As Robert David Sullivan put it, 

After three landslide Republican wins in the 1920s and two landslide wins for Democrat Franklin Delano Roosevelt in the 1930s, the G.O.P. backbone began to take shape in the middle of the United States. Wendell Willkie—the last major-party nominee without experience as a government or military leader—appealed to rural counties in landlocked states resistant to F.D.R.’s New Deal policies. And dozens of these counties have remained resistant to the Democrats, abandoning their earlier flirtations with “prairie populists” like Democrat William Jennings Bryan.

Now, Sullivan is incorrect about why exactly 'rural counties in landlocked states' abandoned the Democracy, as analysed earlier in this post. (It's also worth noting, since Sullivan is implying that 1940, unlike 1924, was not a 'landslide', that FDR got 54.74% of the national vote in 1940, as compared to Coolidge's 54.04% in 1924.) However, it is true that, of the nine states that haven't voted Republican since 1964, four are actually states that haven't voted Republican since 1936 save in 1964 (and in 1964, they all voted to the right of the country). (Before 2008, these four states would have been joined by Indiana.) No remotely similar streak of any particular state voting Republican was begun in either 1944 or 1948. These states are not particularly large, but they have represented a consistently reliable core base for the GOP for over 3/4 of a century now--not only did they all vote for Trump in 2016, but they were also all 'fornia' states for him (and in 2020, he still won all of them by better than 1.25:1). 

Considering Iowa's political past prior to 1932, the only state that would really seem surprising that Willkie carried was Colorado--and Colorado had already voted to the right of the country in both 1932 and 1936 (albeit only slightly so in 1936). This was, perhaps, an indication that, by this time, an important political distinction had already arisen between the 'Mountain-Plains states' and the Pacific Slope states, one that perhaps hadn't existed in the time of Bryan and Wilson.